Comparing Large-Screen-Display Specifications

Specification sheets for large-screen displays list many parameters, but many of them are misleading and not all of them matter to end users. So, more than ever before, it is important to understand the key display specifications and how they differ among technologies.

by Raymond M. Soneira

WE ARE in a renaissance of display technologies. Ten years ago, the cathode-ray tube (CRT) was the dominant display technology. Today, CRT, liquid-crystal-display (LCD), plasma-display-panel (PDP), digital-light-processing (DLP) [using Texas Instruments' Digital Micromirror Device (DMD)], and liquid-crystal–on–silicon (LCOS) technologies are mature and mainstream, with many more trying to emerge from the development laboratory and gain significant market share. How do these technologies differ in performance, and which one should a consumer purchase, particularly for large-screen television?

To provide some substantive answers, my company, DisplayMate Technologies Corp., performed an in-depth comparison of these different display technologies to analyze the relative strengths and weaknesses of each. Direct-view and rear-projection units were included, but front projectors were excluded because they require different measurement and evaluation criteria. We chose the top performer in each technology category from the 2004 DisplayMate Best Video Hardware Guide, which includes our selections for the best video hardware in 40 categories (see www.displaymate.com). The selections are a 40-in. direct-view LCD [NEC LCD4000, Fig. 1(a)], a 61-in. plasma display [NEC 61XM2, Fig. 1(b)], a 50-in. DLP rear projector [Optoma RD-50, Fig. 1(c)], and a much smaller 19-in.-CRT professional high-definition studio monitor [Sony PVM-20L5, Fig. 1(d)] which was used as the reference standard for color and gray-scale accuracy.

We considered including LCOS in this comparison, since it is another upcoming display technology that works as a reflective-mode LCD. However, given the present instability in that segment – Philips and Intel pulled out in October 2004 – we thought that only JVC's D-ILA could be classified as mature and mainstream. It has been used in JVC's front projectors since 1998, but only became available in a rear-projection version after our testing was completed.

This article is a comparison of four different display technologies and not a review of the specific displays listed above. By choosing a top-performing product in each technology category, we are effectively examining the state of the art (as of mid-2004) of each technology.

How We Tested

We carefully set up, tested, and evaluated all of the display technologies at the same time, under identical conditions, using the same procedures. The displays were set up side by side for simultaneous viewing in a completely dark laboratory treated with black felt to eliminate reflections. We used computer and video-based test patterns, plus DVD, television, and computer applications.

For DVI and component video HD signals, we used an ATI Radeon 9800 Pro with an ATI HDTV Component Video Adapter, which provides high-quality computer-generated 720p and 1080icomponent video outputs. This allowed us to generate HD test patterns from our own DisplayMate test and evaluation software for the television video inputs. Our reference standard was the Sony Professional PVM-20L5 Multi-Format High-Definition studio monitor, which was carefully calibrated for testing. Each display was compared to this monitor for color and gray-scale accuracy and overall image quality.

All of the photometry and colorimetry measurements were made with a Konica Minolta CS-1000, which is a high-quality laboratory spectroradiometer with a narrow 1° acceptance angle for light emitted by the display. A narrow acceptance angle is required for accurate measurements for many flat-panel technologies because the display's light distribution can vary with both viewing angle and intensity. The National Institute of Standards and Technology (NIST) and the Video Electronics Standards Association (VESA) specify a maximum acceptance angle of 2° for measuring flat panels. The spectroradiometer and all of the displays (except for the Sony) were generously provided on long-term loan by their manufacturers.

 

Fig__1a_tif (a) Fig__1b_tif (b) Fig__1c_tif (c) Fig__1d_tif(d) NEC, Optoma, Sony

Fig. 1: The tested displays were (a) a 40-in. direct-view LCD (NEC LCD4000), (b) a 61-in. plasma display (NEC 61XM2), (c) a 50-in. DLP rear projector (Optoma RD-50), and (d) a 19-in.-CRT professional high-definition studio monitor (Sony PVM-20L5).

 

Black Level

We started off the comparison with an item that does not get all of the attention that it deserves: the display's ability to produce black. This capability of suppressing light output is a major challenge for all of the display technologies. It is important because a poor black level increases the lower values of the display's intensity scale and introduces errors in both intensity and color throughout the entire lower end of the scale, not just at the very bottom. No display can produce a true black; all displays produce some light in the form of a very dark gray when asked to produce a black. This background light, called black-level luminance, must be added to all of the colors and intensities that the display is asked to produce. Furthermore, if the display is not properly adjusted, the dark background glow will have a color tint instead of appearing neutral gray, and this will add a color cast to the entire lower end of the intensity scale, which is particularly noticeable in dark images.

It is important to know just how close a display can actually get to producing true black. CRTs do extremely well, but the flat panels all have some difficulty in producing black. The actual black level produced by a display is almost never reported in the manufacturer's specification sheets or published reviews, yet for most applications it is actually much more important than peak white luminance, and it is particularly important in multimedia, imaging, photography, home theater, or in any environment with controlled or subdued lighting.

In most displays, the black level is adjusted using a control inappropriately labeled "brightness." Unfortunately, most LCDs lack any form of black-level control; the level is set at the factory. However, many LCDs now have a control labeled "brightness" that instead varies the intensity of the backlight. The NEC LCD4000 is one of a small number of LCDs that actually provide a real black-level control; and it is even labeled "black level."

Black-Level Measurements

The black levels measured with the Konica Minolta CS-1000 Spectroradiometer are listed in Table 1. It measures luminance (which is related to the sensation generally described by human test subjects as "brightness") by matching the human eye's own spectral sensitivity to light of different wavelengths. The measurements were in candelas per square meter (cd/m2). A very sensitive, full-field black-level test pattern was used to set the display's screen to the proper black level. The measurements were made in a completely dark laboratory, so there was no contamination from ambient room lighting.

The CRT won by a very large margin (a factor of about 25). It barely produced any detectable light when set to black. The flat panels all produced a noticeable dark-gray glow for black. The CRT's great black-level advantage is the major reason why it remains the technology of choice for home-theater perfectionists. There are two values listed for an LCD, one when the backlight is set to maximum luminance and the other for minimum luminance. So, for an LCD equipped with a backlight control, a darker black can be achieved if a lower peak white luminance is acceptable. In many instances, that is a very desirable tradeoff.

But how low does the black level really need to be? Home-theater perfectionists insist on a completely dark viewing environment because that is how movie theaters operate. Under these conditions, any noticeable black-level luminance adversely affects image quality and can also be an annoying distraction. But the human visual system employs several light-adaptation mechanisms, and one result is that the threshold for the detection of black-level luminance varies with the average scene luminance over a period that can extend from several seconds up to several minutes for color vision. For typical movie content with varying scene luminance, the eye will be operating at reduced sensitivity and is less likely to notice the black-level luminance in dark scenes, but the eye's sensitivity increases considerably when viewing a movie with predominantly dark scenes, such as Dark City (1998).

Cinema-like Performance

How does the performance of video displays and projectors for home-theater use compare with that of a movie theater? Kodak motion-picture film has maximum densities of roughly 4.0 for the standard Vision Color Print Film 2383 and 5.0 for the high-quality Vision Premier Color Print Film 2393. These densities correspond to dynamic ranges of 10,000 and 100,000, respectively, but production movie prints deliver roughly a factor of 10 less than their specification maximum. CRTs typically have a dynamic range between 10,000 and 30,000, so they can actually perform significantly better than motion-picture film (if they are carefully set up). The best that non-CRT displays and projectors can achieve now is about 3000, so their performance is better than standard-grade motion picture film but well below what the best films can deliver.

The higher the dynamic range, the darker the black level for a given peak luminance. Motion-picture theaters typically operate at a peak luminance between 41 and 75 cd/m2 (SMPTE 196M), which is comparable to that of front projectors but much lower than that of the direct-view and rear-projection displays considered here; therefore, different eye adaptation levels apply.

Color Temperature

Most viewers are aware that there is a whole range of colors that can be accurately referred to as "white." But if we are to have accurate color reproduction, it is necessary to define one or more standard whites, which serve as a point of reference for generating all the other colors. One way to do this is by using the color at which a specially prepared "black body" glows when raised to a specified temperature, which is expressed in degrees Kelvin, or K. Each temperature produces a known spectrum that yields a unique color with specific color coordinates. As the temperature increases, the changing color coordinates trace out a "black-body curve." Whites on this curve typically fall in the range from 5000K (a reddish-white) to 10,000K (a bluish-white).

 


Table 1: Black levels of the Tested Displays
CRT Sony PVM-20L5 LCD NEC LCD4000 Plasma NEC 61XM2 DLP Rear Projection Optoma RD-50
0.01 cd/m2 0.72 cd/m2 Max Backlight 0.42 cd/m2 0.26 cd/m2
0.27 cd/m2 Min Backlight

 

Most computer and television displays come from the factory set to a relatively high color temperature (typically 9300K), which produces a white that has a slight blue cast, similar to "cool white" fluorescent bulbs. For multimedia, photography, and television, the standard color temperature is 6500K, which is roughly the color of natural daylight. For optimum color accuracy, a display for these applications is sometimes set to a white point with the color coordinates of CIE Illuminant D65 (or D6500), which corresponds to the average natural daylight of an overcast sky at noon and includes a blue-sky component added to a black-body spectrum. There are other color standards as well.

If an image is designed or color-balanced at one color temperature and then viewed at a different color temperature, all of the colors in the image will be shifted by varying amounts. For example, reds must be overemphasized in TVs operated at 9300K in order to counteract the blue cast that is imparted to flesh tones. This so-called "red push" introduces other color errors.

In all of our tests, the white point of each display was set as close to D6500 as possible without resorting to any internal service modes. Colors that lie close to, but not exactly on, the black-body curve can be assigned a color-temperature value that produces the closest possible color match to a black body. This is referred to as a correlated color temperature. Below are correlated color-temperature values measured with the Konica Minolta CS-1000 Spectroradiometer and a window test pattern set to peak white (Table 2). The results were all relatively close to D6500, except for the video inputs on the NEC LCD4000, which did not provide any adjustments for the white point.

The color temperature (and chromaticity coordinates) should not change as the gray-scale intensity changes, but it always does so to some degree because of slight differences between the display's red, green, and blue channels. This variation is called color tracking or gray-scale tracking, and one benchmark of a good display is a small variation. All of the displays passed the color-tracking tests quite well.

Peak White Luminance

Under most typical viewing conditions, the display technologies evaluated here all deliver more than enough light for comfortable viewing, so a higher peak luminance is not necessarily better. On the other hand, if bright ambient light cannot be reduced, high brightness may become an important requirement; and phosphor and lamp aging will reduce luminance over time, so some reserve is necessary.

There are NIST/VESA, ANSI, and ITU-R standards for measuring the peak white luminance, but they all have some "wiggle room" that allows the numbers to be exaggerated. Even worse, many manufacturers' specifications sheets do not reference any standard, so they are free to choose their own procedures. Frequently, every control that can increase the light output is turned up to maximum in order to produce a relatively high luminance value that can be reported. Under these conditions, practically all displays will produce very poor image quality. When luminance becomes a significant issue, attention should be paid to only the values measured under identical standard conditions. Press reviews are generally the best source.

The "contrast control" is the primary means of adjusting peak white luminance and the top end of the intensity scale. Despite its name, it does not affect the display's contrast. If it is set too high, two or more of the top-end levels in a gray-scale test pattern will reach peak luminance and merge together. This loss of gray scale is called either "white saturation" (a soft limit for CRTs and LCDs) or "clipping" (a hard limit for plasma displays and DLP rear projectors. In many applications, the display does not need to be operated at peak luminance. In fact, some displays are now so bright that under typical indoor lighting conditions they may cause enough discomfort to compel some viewers to dim them. To reduce peak luminance, the contrast control must be turned down, or, in the case of an LCD, a backlight "brightness control" must be turned down. When lowering the contrast control, the black-level control may need some adjustment because they interact.

Peak-White-Luminance Measurements

In our tests, all of the monitors were set up the same way. First, the white point was set as close to D6500 as possible, the black levels were carefully adjusted as described previously, and then the contrast control was set by using a DisplayMate White Saturation test pattern so that no more than 2% of the gray scale was lost near peak white. For CRTs there are additional requirements for focus and screen regulation, but they did not affect the Sony monitor.

The values obtained with this procedure will generally be lower, and sometimes much lower, than those listed on a display's specification sheet. The luminance levels measured with the Konica Minolta CS-1000 Spectro-radiometer and a Windows test pattern set to peak white are listed in Table 3. The LCD has two entries, which depend on the backlight-intensity setting. At its highest available color-temperature setting of 9023K, the LCD produced a luminance of 471 cd/m2; this is higher than that listed by NEC on their specification sheet, in a step which is both unusual and commendable!

 


Table 2: Correlated Color Temperatures
CRT Sony PVM-20L5
LCD NEC LCD4000
Plasma NEC 61XM2
DLP Rear Projection Optoma RD-50
6480K
6,580K Computer Inputs
6626K
6786K
10,250K Video Inputs

 


Table 3: Luminance Levels (Peak White)
CRT Sony PVM-20L5
LCD NEC LCD4000
Plasma NEC 61XM2
DLP Rear Projection Optoma RD-50
176 cd/m2
428 cd/m2 Max Backlight
212 cd/m2     5% APL
359 cd/m2
 
160 cd/m2 Min Backlight
133 cd/m2   25% APL
 
   
81 cd/m2     50% APL
 
   
53 cd/m2   100% APL

 

The values for the plasma display depend on the average picture level (APL), which is the average intensity level of each of the red, green, and blue subpixels over the entire screen. For example, a full screen of peak white intensity has an APL of 100%, but it is only 33% for pure green because the red and blue subpixels are off.

In our case, APL refers to the percentage of pixels that are set to peak white. When 5% of the pixels are at peak white, the luminance is 212 cd/m2. As the APL increases, power and heat-dissipation restrictions reduce the maximum luminance that can be safely produced, so the display automatically reduces the peak luminance. When 100% of the pixels are at peak white, the luminance is only 53 cd/m2, which requires subdued ambient lighting for good viewing.

For most computer applications, the APL is rather high but in most video applications it is relatively low because the images are generally dimmer and are colored rather than gray or white. This makes plasma displays better suited to video applications.

Dynamic Range and Display Contrast

Dynamic range is simply the ratio of peak white luminance to black-level luminance that a display can produce. The values are measured separately, one screen for peak white and the other for the black level. This is frequently referred to as "contrast," "full-field contrast," or "full on/off contrast," but the term contrast should really be reserved for measurements performed on a single image, not on different screens. The ratio of the peak white to black-level luminance tells us the maximum range of luminance that the displaycan produce. Thus, dynamic range is especially important in imaging and home-theater appli-cations because bright/day scenes and dark/night scenes must both be rendered accurately.

The higher the dynamic range, the better the display will be able to reproduce wide differences in scene luminance. Note that a high dynamic range will also yield a dark black level unless the peak luminance is very high. The ratios calculated from the peak-white and black-level values measured above are listed in Table 4.

Again, the CRT won by a large margin. (Using a sensitive photometer, we have measured dynamic-range values as high as 36,500 for a CRT.) The CRT's enormous lead in dynamic range is another major reason why it remains the technology of choice for home-theater perfectionists. There are four values for the plasma display, depending on the APL of the peak-white field. Note that there is only a single value listed for the LCD because the peak-white and black-level values track exactly with the backlight intensity.

Among the flat panels, the DLP rear projector led by more than a factor of 2, and the plasma display trailed the LCD by 15% for low APL and by a much larger factor for high APL. Remember that these values were measured in a completely dark laboratory. Ambient room lighting would necessarily decrease the above values because the black levels would be higher.

If the peak luminance is lowered by means of the contrast control, the dynamic range will also be reduced because the black-level luminance generally does not change. This is a major advantage for LCDs having a backlight control and projectors having an iris aperture control. Their dynamic range remains constant because the black-level luminance decreases together with the peak luminance.

Display contrast is another widely advertised specification, but this number fluctuates more than any other specification. It is supposed to represent the ratio of the brightest white to the darkest black that a display can produce within an image. Because internal reflections within a display or display optics cause light from the bright areas of the image to bleed into the dark areas, they cannot get as dark as the black levels listed above. Thus, the display contrast is always less than the dynamic range. If the display's contrast falls too low, then images will appear washed out. Unless a standard such as ANSI is listed next to the contrast specification, it is most likely some form of dynamic range.

Contrast Measurements and Interpretation

A standard way to measure display contrast is to use a black-and-white checkerboard test pattern and measure the luminance at the center of the white blocks and then that of the black blocks. The smaller the blocks, the greater the bleed, resulting in lower contrast values. We have done this for a 4 x 4 checkerboard pattern, which is a standard pattern, and then for a much finer 9 x 9 checkerboard to see how much more the contrast falls when the area of the blocks is further reduced by a factor of 5.

This measurement is tricky because a similar contamination effect, called veiling glare, also affects the measuring instrument. We used heavy black-felt masks to eliminate this common source of error in contrast measurements. All of the displays had their controls carefully adjusted as described previously. The measurements were made in a completely dark laboratory, so there was no contamination from ambient room lighting (Table 5).

Comparing the 4 x 4 checkerboard values with dynamic range, we see that the CRT value fell the most, by a factor of 80, because of multiple reflections within its thick glass faceplate.

The DLP rear-projector value fell by a factor of 4, primarily due to reflections within the rear-projection optics. The LCD value decreased by only 2% because the glass is thin and multiple reflections are heavily absorbed. For similar reasons, the plasma-display value also showed a relatively small 6% decrease from the dynamic-range values.

For the much finer 9 x 9 checkerboard, there was a comparatively smaller decrease despite the fact that the blocks each had one-fifth the area of the 4 x 4 checkerboard pattern.

The term "contrast" has been applied in so many ways that its meaning is no longer clear. Almost all manufacturers' "contrast" specifications actually refer to the display's dynamic range rather than anything indicative of the luminance ratios that will be generated for an image by the display. Checkerboard display contrast certainly falls within the definition of contrast that we have been discussing. However, it generally does not correspond well with the human eye's own sense of visual contrast.

 


Table 4: Dynamic Range (Peak White Divided by Black Level, Full Screen)

CRT Sony PVM-20L5 LCD NEC LCD4000 Plasma NEC 61XM2 DLP Rear Projection Optoma RD-50
17,600 595 505       5% APL 1,381
317     25% APL
193     50% APL
126   100% APL

 

In particular, the human eye does not really notice the large differences in display contrast that were measured. Side by side, the checkerboard patterns on all of the displays appeared to have roughly the same visual contrast, even though the instrumentation indicated otherwise. The human eye can detect that there are differences, but they appear to be small differences, instead of the factor of 3 measured for the 4 x 4 checkerboard and the factor of 8 measured for the 9 x 9 checkerboard pattern.

This has much more to do with human visual perception than with optics. The human eye is, after all, not a camera or an instrument, but an image-processing system that is designed to extract visual information together with the brain, which supplies the processing and interpretation. It seems that on these scales the brain indicates that there are large luminance differences between the adjacent bright and dark checkerboard blocks, but is less concerned with their precise ratio because there is no perceptual content involved. There is no question that if the checkerboard contrast falls too low, the human eye will at some point take full notice of the effect, but these displays did not trigger that response.

It is an entirely different story for the smaller scale used in fine text and graphics. For black text on a white background, the human eye immediately notices that characters on the CRT show up as light gray on white instead of very dark gray on white for the flat panels, and it is definitely harder to read fine text on a CRT than on any of the flat panels. The differences in display contrast are clearly significant in this case.

The optics in front and rear projectors also has a major impact on display contrast because each element in the light path scatters a small fraction of the light that reflects off or passes through it. That is why the rear-projection DLP experienced a significant decrease from the dynamic-range value.

We have seen that measuring checkerboard display contrast is tricky and its interpretation is often ambiguous and misleading, so its usefulness is limited. We need a measure of contrast that has a better correspondence with the human eye's own sense of visual contrast. There is such a measure, called image contrast, which depends on the details of the gray scale, and particularly on a widely misunderstood parameter called gamma. So let us take a closer look at them.

Gray Scale

We have discussed and measured the extremes of display luminance: the black level and peak white intensity. Now we will examine all of the intensities in between, which is referred to as the display's gray scale, or, more technically, its transfer function. This is the signature of a display; it is what gives the display its own unique look and performance characteristics.

While each display technology has its own native gray scale – the transfer characteristic of the display device – signal-processing electronics within the display modifies this to produce the gray scale that we actually see (and measure). One of the reasons this is necessary is that there must be a standard gray scale so that images will be accurately reproduced on any display or display technology. The accepted standard is the CRT's own native gray scale because the CRT was until recently the dominant display technology which new technologies had to mimic if they were to be accepted and because the CRT's native gray scale is very close to the theoretical ideal.

"Gray scale" is actually an unfortunate choice of words because the relationship in question describes the brightness scale for all colors, not just the grays. That is because the same functional form must apply to each of the red, green, and blue display primaries so that the display's color balance does not vary with brightness.

Before going any further, we must define exactly what is meant by a gray scale. It is the luminance or amount of visible light that a display produces for a given level of input signal, and applies to every pixel in the image. For instance, a maximum signal produces peak white and a zero signal produces the closest approximation to black that the display can provide.

We measured the gray scale using a set of our own DisplayMate for Windows test patterns together with the Konica Minolta CS-1000 Spectroradiometer. When we increase the signal from zero to maximum, the display luminance increases in a particular way that we can measure and then plot on a graph. This graph of luminance vs. signal intensity is called the display's gray scale.

The input signal can be specified in many different but equivalent ways. For computers, it is generally on a scale of 0–255, with 0 for black and 255 for peak white. For most digital video, it is 16 for black to 235 for peak white; and for analog video, it is generally specified in IRE units, from either 0 or 7.5 for black to 100 for peak white. To simplify matters, we will describe the input-signal intensity level on a scale from 0% for black to 100% for peak white.

What Is Gamma?

Gamma is a popular yet widely misunderstood number that describes the steepness of a display's gray scale as it increases from black to peak white. The gray scale is not linear but is instead logarithmic because that is how standard CRTs behave and because it corresponds well with the human eye's own logarithmic response.

 


Table 5: Contrast (Checkerboard)
CRT Sony PVM-20L5 LCD NEC LCD4000 Plasma NEC 61XM2 DLP Rear Projection Optoma RD-50
4 x 4 Checkerboard 219 586 475     5% APL 332
         Contrast 305   25% APL
188   50% APL
124   High APL
9 x 9 Checkerboard 75 577 449     5% APL 274
         Contrast 294   25% APL
184   50% APL
122   High APL

Note: The checkerboard pattern has a 50% APL. For the plasma displays, values for the other APLs were calculated by applying the same factors for the light bleed to the peak white luminance values. The "High APL" entry uses the values for 100% APL.

 

Figures 2 and 3 are logarithmic plots of the gray scales of our test displays as measured with the Konica Minolta CS-1000 Spectro-radiometer and a set of DisplayMate for Windows test patterns. Because both scales are logarithmic, they are log-log plots and render power-law plots as straight lines. All of the display controls had to be adjusted very carefully for these measurements, especially the black level.

It can be seen that the gray scales are all reasonably close to straight lines on a log-log graph, but have different slopes. The logarithmic plot emphasizes the dimmest parts of the gray scale, which parallels the human eye's own extended sensitivity to dark content. (On a linear graph, the dimmest parts of the gray scale would be compressed into a tiny area at the lower left corner, so they would be virtually invisible and the important behavior at the dim end lost altogether.)

The LCD has the steepest gray scale and the plasma display the shallowest (Fig. 2). This explains why one FPD technology sometimes appears to be the brightest and at other times a different one does. Above a signal intensity of 55%, the LCD is the brightest and the plasma display the dimmest, but below a signal intensity of 20% the plasma display is the brightest and the LCD is the dimmest. Between 20 and 55%, the DLP rear projector is the brightest. Thus each of the flat-panel displays gets a turn at being number one. So the relative brightness of the flat panels actually depends on the image content.

Figure 3 shows the same plot as Fig. 2, but now the peak luminances are all equalized to a common 100%. This allows us to compare the relative behavior of the gray scales directly because they now all have the same ratios compared to peak brightness. It is now easy to see why the displays look different. There is a considerable variation in the steepness of the gray scales and the differences diverge significantly at lower signal intensities. At a signal intensity of 20%, there is almost a 3:1 difference in the relative brightness between the plasma display and the LCD.

Gamma is the numerical value of the slope (steepness) of the gray scale when plotted on a logarithmic log-log graph. While a gamma of 3.0 might be considered optimum based on the human eye's specific power-law response, it is more important to have a standard gamma value defined. Television-, DVD-, Web-, and computer-based photographic content are generally color-balanced on professional CRT studio monitors that are electronically adjusted to a standard gamma of 2.20 because the most accurate images will be obtained at this value. We determined the gammas of the test displays from the log-log plots in the most important region of a signal intensity, from 100 to 30% (Table 6). Below 30%, the slopes start varying somewhat, which means the gammas will also vary at the very dark end.

The gamma for the Sony CRT agrees perfectly with the 2.20 standard value. The LCD has a gamma greater than the 2.20 standard, the DLP rear projector has less, and the plasma display has much less. The plasma display and DLP rear projector both provide a choice among several gamma selections. We chose the steepest gamma available for each, which provides the closest agreement with the 2.20 standard. How much of a difference do these different gray scales and values of gamma make, and how do they affect the appearance of an image?

 

fig2_tif

Fig. 2: Luminance in cd/m2 for each of the four displays tested is presented as a function of the signal-intensity level expressed as a percentage of maximum. The open symbols plotted on the graph are the measured data points. The values for 100% are identical to the peak-white-luminance values listed in Table 3, with a backlight set to maximum for the LCD and a 5% APL for the plasma display.

 

fig_3_tif

Fig. 3: This is a replotting of the data in Fig. 2, with the peak luminances now equalized to a common 100%. The relative behavior of the gray scales can be directly compared because they now all have the same ratios compared to peak brightness and it is easy to see why the displays look different. At a signal intensity of 20%, there is almost a 3:1 difference in the relative brightness between the PDP and the LCD.

 

Gamma has a major effect on image brightness, contrast, hue, and color saturation. To explore this, let us examine the differences between the LCD, which has the highest gamma, and the plasma display, which has the lowest. To see this on the graph in Fig. 3, the luminance of the LCD and plasma display at a signal intensity of 50 and 25% must be compared. At 50%, the plasma display has a luminance that is 1.3 times that of the LCD. At 25%, that luminance ratio increases to 2.1.

Gamma and Image Characteristics

Contrast. The higher the gamma, the faster the luminance decreases with signal intensity. Consider a black-and-white photograph. The luminance ratio of the bright content to the dark content will be considerably greater for the LCD than for the plasma display. These ratios of luminance are actually just contrast ratios. For example, at an intensity level of 25%, the LCD will have a contrast that is 2.1 times that of the plasma display. Overall, the LCD image will appear to have a higher contrast than that of a standard 2.20-gamma display, and the plasma display will have a lower contrast. So, gamma is the determining factor in visual contrast for images shown on a display. A control that varies gamma would function as a true contrast control.

Image brightness. Again, the higher the gamma, the faster the brightness decreases with signal intensity. Since most images have a wide range of intensities, displays with a higher gamma will appear darker and those with a lower gamma will appear brighter. Given the industry's emphasis on brightness, it is not surprising to find a bias towards lower gamma values.

Hue. When the primary colors are combined to produce color mixtures in an image, different gammas result in different luminance values for the primaries, which produces varying hues and brightness in the resulting colors. For example, when mixing red and green in the ratio of two parts red to one part green, which produces a brown, the green will be 1.3 times brighter on the plasma display than on the LCD, so the browns will be different. We verified this visually; the brown was noticeably redder on the LCD than on the plasma display. From this, we see that image hues are significantly affected by gamma, including flesh tones. Most users will tweak the tint and other controls to make the flesh tone come out right. But every tweak that is used to compensate for a display-parameter error leads to a progression of inaccuracies that add up.

Color saturation. Color saturation is also affected by gamma in the same way as hue, except that all three primary colors are involved. The primary color with the lowest signal intensity in any color mixture determines the saturation of the resulting color because it is perceived as combining with equal intensities of the two other primaries to produce a low intensity shade of white (dark gray). This washes out the appearance of the color mixture into a lower-saturation pastel. Since gamma has the greatest effect on the dimmest primary color, the brightness of this white (gray) component varies significantly. For example, a mixture of 75% red, 50% green, and 25% blue is perceived as a red–green mixture having a 25% white (gray) component.

This white component will be 2.1 times brighter on the plasma display than on the LCD, so the color will have a much lower saturation on the plasma display than on the LCD. As a result, color saturation is significantly affected by gamma. Higher gammas increase color saturation and lower gammas decrease color saturation. The display's color-saturation control can be used to reduce the saturation error resulting from a non-standard gamma, but it cannot eliminate it because the saturation error varies with signal intensity. Again, users will tweak the saturation control to make the flesh tones come out right, but all of the other colors will be modified at the same time.

So, a standard gamma is necessary to obtain accurate color saturation at all intensities. However, a gamma higher than the standard 2.20 can be used with flat-panel displays to counteract the reduced color saturation that is due to an elevated black-level luminance.

Variations in gamma. In principle, the gray scales should appear as perfectly straight lines (power laws) in Figs. 2 and 3. Otherwise, the gamma of a display will vary with signal intensity, and so will all of the characteristics discussed above (as well as saturation, which we have not discussed). In the brown example we described, a display will produce different browns at different signal intensities if the gray scale deviates from a straight line in a log-log plot. Note that the CRT's gray scale is almost perfectly straight.

Gamma Control

Although having a standard gamma of 2.20 is important, high-quality displays should have a gamma control to allow image contrast to be adjusted in response to variations in the source material, ambient lighting, and individual preferences. (Remember that the control labeled "contrast" on most displays actually controls image brightness and does not affect image contrast.)

To properly adjust image contrast, the gamma control must vary the logarithmic slope of the gray scale. If the nominal value is 2.20, then a range of plus or minus 0.4 (with steps no greater than 0.1) should be sufficient to accommodate most source-material variations and individual preferences. The lower gamma values are good for improving source material that has too much contrast or for boosting overall image brightness. The higher gamma values are good for boosting source material that has insufficient contrast or for improving image contrast that has been reduced by bright ambient lighting washing out the screen.

Higher contrast values may also be required in very dark ambient lighting conditions because the human eye tends to reduce visual contrast in those situations. So the gamma range must be extended by an additional 0.4 on the plus side to accommodate these effects. A gamma range of 1.8–3.0 should cover just about all situations.

 


Table 6: Gammas of the Test Displays
CRT Sony PVM-20L5
LCD NEC LCD4000
Plasma NEC 61XM2
DLP Rear Projection Optoma RD-50
2.20
2.32
2.02
2.09

 

Gamma Interpretation

There are many reasons why displays have different gammas. While each technology has its own native transfer characteristic (or gamma), the display's signal-processing electronics modifies it in order to obtain the desired gray scale as accurately as possible. In particular, current CRTs typically have a native gamma in the range of 2.3–2.6, so a gamma of 2.20 found in Sony and Ikegami CRT studio monitors is actually the result of signal processing. DLP rear projectors and plasma displays have a native gamma of 1.0, and LCDs have a variable native gamma that results from an S-shaped transfer characteristic. So signal processing plays an important role in generating a display's gray scale.

One consequence of this is that the specific gamma values we have measured for our test displays apply only to these particular models and are not inherent to their particular technologies. But the behaviors that we have seen here are not arbitrary or accidental; in fact, they have been carefully chosen by their manufacturers to complement each display technology's strengths and weaknesses.

For example, the LCD is currently optimized for computer applications, where signal intensities are frequently near peak. The steep gray scale produces bright, high-contrast images with high color saturation. The plasma display is optimized for video applications, which have much lower signal intensities. The less-steep gray scale helps deliver a brighter image. The DLPrear projector is relatively bright, so it can afford to use a steeper gray scale at low intensities to enhance visual contrast and color saturation near its black level. •

 


Raymond M. Soneira is President of DisplayMate Technologies Corp., P.O. Box 550, Amherst, NH 03031; telephone 603/672-8500, fax 603/672-8640, e-mail: rmsoneira@ displaymate.com. The author thanks Edward F. Kelley (NIST), Craig Verbeck (Pixelworks), John P. Pytlak (Eastman Kodak Co.), and Alan Keil (Ikegami Electronics) for enlightening technical discussions and valuable information. Particular thanks go to the Konica Minolta Instrument Systems Division for providing editorial loaner instruments whenever and wherever they have been required and for providing the CS-1000 Spectroradiometer on a long-term loan for this project.